Optimization Method of Dynamic Monitoring of Groundwater Level

Abstract

The invention relates to an optimization method for dynamic monitoring of groundwater level. The method comprises the following steps: in a groundwater monitoring area, a plurality of sensors are evenly arranged, including water pressure, temperature, humidity multi-dimensional parameter monitoring; the sensors form a grid monitoring system through a wireless network; the multi-dimensional data collected by the sensor array is input into a data fusion module, and the data fusion algorithm is used to perform collaborative processing on various data to filter out noise and abnormal values; a unified water level change model is constructed to make data of different parameters complement each other; the weight in the fusion algorithm is dynamically adjusted according to historical data and new measurement values ​​on a regular basis; in the monitoring area, a spatiotemporal analysis model of groundwater level change is constructed based on the dimensions of time and space to dig out the periodicity and regularity of water level fluctuation; a time series analysis algorithm including ARIMA or LSTM is used to perform trend analysis on historical water level data to generate short-term and long-term water level predictions.

Description

Technical Field

[0001] The invention relates to an optimization method for dynamic monitoring, in particular to an optimization method for dynamic monitoring of groundwater level. Background Art

[0002] At present, China's invention patent 202410799927X proposes a method for dynamic monitoring of groundwater levels based on wireless sensor networks, which specifically includes setting up water level observation holes in the measured area, laying sand filter layers, installing wireless sensor fulcrums, backfilling soil and setting up wireless gateways. This method realizes the monitoring of groundwater levels through the deployment of wireless sensor networks and real-time analysis of signal strength changes, thereby improving the real-time and automation of monitoring. However, although this method is innovative and has practical application value in some aspects, it still has many shortcomings in the overall optimization of dynamic monitoring and management of groundwater levels.

[0003] First, this method mainly relies on the signal strength change of the wireless sensor network to reflect the dynamic change of the groundwater level. Its principle is to indirectly measure the water level by analyzing the signal attenuation. The main problem with this method is that the measurement accuracy is not high. Especially in the face of complex groundwater environment, the interference, noise, attenuation characteristics of the signal and the physical installation conditions of the sensor may have a great impact on the signal strength, which in turn limits the accuracy of the water level data. In addition, the attenuation of the wireless signal is not only affected by the change of the water level, but also by various factors such as soil type, humidity, sensor spacing, and the density of the backfill soil. This requires frequent calibration and debugging of the system to ensure the reliability of the data, which increases the maintenance difficulty and cost in actual operation. Secondly, although this method improves the real-time monitoring of groundwater level, its monitoring data processing and analysis capabilities are relatively limited. It is mainly based on the qualitative analysis of signal strength changes and lacks an in-depth interpretation of water level changes in complex environments. Since the groundwater level is judged only by the signal change of a single monitoring point, this method cannot fully capture the spatiotemporal dynamic characteristics of the groundwater level, the spatial correlation analysis between monitoring points is insufficient, and the collaborative processing capability of multi-point data is lacking. Groundwater level changes often have strong temporal and spatial correlations and are affected by multiple factors such as precipitation, evaporation, and geological structure. Traditional signal strength monitoring methods are difficult to analyze and predict dynamic changes in water levels from multiple dimensions and scales. Therefore, this method cannot effectively explore the periodicity and regularity of water level changes and is difficult to provide in-depth data support for the optimal management of groundwater resources.

[0004] Thirdly, the patented method has weak detection and feedback capabilities for abnormal water level fluctuations. Although it is capable of real-time monitoring of water level changes, when abnormal fluctuations occur, the alarm is triggered only by changes in signal strength. This alarm mechanism based on a single factor is susceptible to false alarms or missed alarms, especially when the environment changes greatly or the signal is subject to much interference. Due to the lack of comprehensive environmental variable analysis and feedback control mechanisms, this method is difficult to respond quickly and adjust when abnormal water level fluctuations occur, which may lead to excessive extraction or insufficient replenishment of groundwater resources, affecting the sustainable use of water resources. In addition, this method does not combine advanced technologies such as spatial coupling analysis or deep learning to perform time-series prediction and trend analysis of water levels, and cannot provide early warning of potential water resource crises and provide effective regulatory recommendations for decision makers. Summary of the invention

[0005] The purpose of the present invention is to provide an optimized method for dynamic monitoring of groundwater level, thereby solving some of the drawbacks pointed out in the background technology.

[0006] The present invention solves the above-mentioned technical problems by adopting the following technical solution, which includes the following steps:

[0007] S1. Groundwater level monitoring based on sensor array:

[0008] S1.1. Multiple sensors are evenly distributed in the groundwater monitoring area, including water pressure, temperature, humidity and other multi-dimensional parameter monitoring;

[0009] S1.2, the sensors form a grid monitoring system through a wireless network;

[0010] S2. Multi-parameter collaborative groundwater level data fusion:

[0011] S2.1, input the multi-dimensional data collected by the sensor array into the data fusion module, use the data fusion algorithm to coordinate the data and filter out noise and outliers;

[0012] S2.2, build a unified water level change model so that data of different parameters can complement each other;

[0013] S2.3, regularly adjust the weights in the fusion algorithm based on historical data and new measurements;

[0014] S3. Prediction of groundwater level change trend based on spatiotemporal analysis:

[0015] S3.1. In the monitoring area, based on the dimensions of time and space, a spatiotemporal analysis model of groundwater level changes is constructed to explore the periodicity and regularity of water level fluctuations;

[0016] S3.2, use time series analysis algorithms including ARIMA or LSTM to perform trend analysis on historical water level data and generate short-term and long-term water level forecasts;

[0017] S3.3. Dynamically adjust the parameters of the spatiotemporal analysis model to adapt to changes in different environmental factors including seasonal precipitation and geological changes;

[0018] S4. Groundwater level anomaly monitoring and adjustment mechanism based on dynamic feedback:

[0019] S4.1. Set dynamic thresholds based on water level data predicted by spatiotemporal analysis, monitor and compare actual data with predicted values ​​in real time, and detect abnormal fluctuations;

[0020] S4.2. Use feedback control algorithms to make immediate adjustments to abnormal situations and increase monitoring intensity in abnormal areas by adjusting monitoring frequency and range.

[0021] Furthermore, the spatiotemporal analysis model construction method includes: using a long short-term memory network (LSTM) to train historical water level data and environmental variables, capturing the nonlinear dynamic characteristics of the time series, and predicting future water level changes; combining Fourier transform to extract periodic characteristics and identify short-term fluctuations and long-term trends in water level changes; and exploring the inherent laws of water level changes:

[0022]

[0023] in, is the predicted future water level change value, which represents the water level prediction result after k time steps in the future; X(t) is the input data matrix at the current moment, which contains the historical water level data of all monitoring points; θ is the parameter set of the LSTM model, including weights and biases; α(t′) represents the time weight function of the input feature, reflecting the importance of the feature at different time points; g(τ,β) is a time-dependent dynamic learning function, which describes the time dependence of water level change, and the parameter β controls the shape and sensitivity of the function;

[0024] The nonlinear trend of water level change is captured by internal integration of the function g(τ,β); the change of water level within k time steps is comprehensively predicted by external integration; Fourier transform is used to transform the water level change in time domain into frequency domain analysis, identify the periodic characteristics P of water level fluctuations, and form a comprehensive description of short-term and long-term trends.

[0025] Furthermore, the spatiotemporal analysis model construction method includes: constructing a spatial coupling model, combining the geographic information system GIS and spatial statistical methods to perform spatial correlation analysis on water level data at different monitoring points, quantifying the coupling degree of water level changes in different regions, and identifying areas of related or independent changes; optimizing the scheduling strategy of groundwater resources by constructing a multi-objective optimization model, and the spatial coupling and scheduling optimization adopt the formula:

[0026]

[0027] Among them, Q(t) represents the amount of groundwater extracted, representing the extraction strategy at time t; D(t) is the demand for water resources, representing the user's water demand; C(t) is the comprehensive cost function, including the energy consumption and resource loss costs during the extraction and recharge process; λ1 and λ2 are optimization weight coefficients used to balance the relationship between extraction demand and cost; N is the number of monitoring points; w ij Represents the elements of the spatial weight matrix, reflecting the spatial correlation between monitoring points i and j; X i and X j are the water level values ​​of adjacent monitoring points, respectively, which characterize the impact of spatial coupling on water level changes.

[0028] Furthermore, the spatiotemporal analysis model construction method includes: introducing a feedback control mechanism, using real-time monitoring data and prediction results to dynamically adjust the extraction and replenishment strategies; combining genetic algorithm optimization control strategies to control the dynamic balance of multiple objectives, and the adaptive control module adjusts the control parameters to form a closed-loop optimization system when receiving new water level change information, and the feedback control and optimization functions:

[0029]

[0030] in, is the adjustment rate of the scheduling model parameters, indicating the speed of parameter change over time; J(Q, D, θ) is the objective function, which comprehensively considers the relationship between extraction, demand and control parameters, and reflects the degree of optimization of the current scheduling strategy; are the partial derivatives of the objective function with respect to the extraction quantity, demand quantity and parameters, respectively, which measure the impact of the variables on the scheduling strategy; γ is the learning rate, which controls the adjustment speed.

[0031] Furthermore, the method for detecting abnormal fluctuations includes: dynamically calculating the threshold of each monitoring point through spatiotemporal analysis, combining historical water level data and environmental variables including precipitation and evaporation; the dynamic threshold reflects the fluctuation of the current water level and is adjusted according to seasonal changes, climate conditions and historical fluctuation patterns:

[0032]

[0033] Among them, T(t) represents the dynamic threshold at time t; μ(W(t)) is the mean of the historical water level data, reflecting the central trend of the water level; σ(W(t)) is the standard deviation, indicating the degree of water level fluctuation; k is the control coefficient, which is used to adjust the sensitivity of the dynamic threshold; α is the influencing factor of periodic fluctuations, which determines the periodic change of water level over time; P represents the analysis period; φ is the phase, which controls the starting position of the fluctuation; and β and γ represent the influencing factors of precipitation and evaporation on the threshold, respectively.

[0034] Furthermore, the method for detecting abnormal fluctuations includes: continuously comparing the real-time water level data with the predicted dynamic threshold; identifying abnormal fluctuations of the water level, triggering an alarm and taking corresponding measures; reducing false alarms by setting a tolerance range:

[0035]

[0036] Among them, Alert indicates the alarm state. When the real-time water level W real When (t) exceeds the dynamic threshold T(t) plus the tolerance ∈ or minus the tolerance ∈, the alarm value is 1, indicating that an alarm is triggered; otherwise it is 0, indicating a normal state.

[0037] Furthermore, the method for detecting abnormal fluctuations includes: identifying the correlation between different monitoring points by performing water level time series prediction and combining spatial coupling analysis; based on the data of a single monitoring point, covering the water level changes of other points in the area; and predicting the results by weighting historical data:

[0038]

[0039] in, is the predicted value of future water level; g(·) is the mapping function of the deep learning model, which is responsible for integrating input data and generating output; S(t′) represents the factors related to spatial coupling; θ is the parameter set of the model, including network weights and biases; and the integral symbol represents the weighted average of historical data.

[0040] The beneficial effects of the present invention are mainly reflected in the following aspects:

[0041] By introducing spatiotemporal analysis models, deep learning predictions, and spatial coupling analysis, the threshold of each monitoring point is calculated to identify abnormal water level fluctuations in real time. Compared with the traditional fixed threshold method, this high-precision dynamic monitoring method significantly improves the sensitivity and response speed to water level changes. It can quickly trigger alarms and take corresponding measures when abnormal situations occur to prevent further deterioration of groundwater resources.

[0042] Through the optimization of feedback control mechanism and genetic algorithm, the extraction and recharge strategies can be dynamically adjusted according to real-time monitoring data and prediction results. When receiving new water level change information, the adaptive control module automatically adjusts the control parameters to form a closed-loop optimization system; and by setting the tolerance range and optimizing the alarm mechanism, the false alarm rate of the monitoring system is effectively reduced. This optimization not only reduces the unnecessary waste of resources caused by false alarms, but also improves the overall stability and management efficiency of the monitoring system, allowing water resource management personnel to focus more on dealing with real abnormal situations. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 This is a flow chart of the optimization method for dynamic monitoring of groundwater level of the present invention.

[0044] Figure 2 The present invention is a flow chart of the method for constructing a spatiotemporal analysis model.

[0045] Figure 3 The present invention is a flow chart of the method for detecting abnormal fluctuations. DETAILED DESCRIPTION

[0046] The specific implementation modes of the present invention will be described in detail below in conjunction with the accompanying drawings.

[0047] Combination Figure 1 The process of groundwater level monitoring based on sensor array is as follows: first, multiple sensors are evenly arranged in the groundwater monitoring area. These sensors can monitor multi-dimensional parameters such as water pressure, temperature and humidity. This multi-dimensional monitoring method can provide more comprehensive and accurate groundwater level information, because the water level is not only affected by groundwater flow, but also by environmental factors such as temperature and humidity. By collecting these data at the same time, we can better understand the dynamic changes of groundwater and its driving factors. Secondly, the sensors form a grid monitoring system through a wireless network. This grid structure enables the sensors to be interconnected in real time and transmit data to the central processing unit.

[0048] The introduction of wireless networks not only improves the flexibility and scalability of the monitoring system, but also reduces wiring costs and construction complexity, ensuring that monitoring data can be transmitted efficiently, facilitating real-time analysis and decision support. In addition, the grid monitoring system can cover a wider area, ensuring the uniformity and representativeness of monitoring data, thereby providing more reliable basic data for dynamic monitoring of groundwater levels and promoting scientific management and effective utilization of groundwater resources.

[0049] The multi-parameter collaborative groundwater level data fusion first inputs the multi-dimensional data collected by the sensor array into the data fusion module. This step is crucial because the data monitored by different sensors may have different characteristics and errors, and directly using them for analysis may lead to inaccurate results. By using the data fusion algorithm, various data can be processed collaboratively, noise and outliers can be filtered out, and the quality and reliability of the data can be improved.

[0050] This processing not only eliminates random errors, but also identifies systematic deviations, thereby ensuring that the basic data for subsequent analysis is accurate. Next, a unified water level change model is constructed so that the data of different parameters complement each other. This means combining different types of data such as water pressure, temperature, and humidity to form a comprehensive model to more comprehensively reflect the changes in groundwater levels. In this way, the model can capture the interrelationships and influencing mechanisms between various parameters, further improving the accuracy and reliability of predictions. Finally, the weights in the fusion algorithm are dynamically adjusted based on historical data and new measurements on a regular basis. This is to ensure that the data fusion process can adapt to changes in the environment and fluctuations in sensor performance, and to update the weights in a timely manner to make them more in line with the current actual situation, thereby enhancing the adaptability and accuracy of the model.

[0051] Based on the prediction of groundwater level change trend based on spatiotemporal analysis, in the monitoring area, firstly, a spatiotemporal analysis model of groundwater level change is constructed based on the dimensions of time and space. The establishment of this model aims to deeply explore the periodicity and regularity of water level fluctuations, and identify potential patterns in water level changes, such as daily changes, seasonal changes and long-term trends, by analyzing historical data. This spatiotemporal analysis can not only reveal the basic laws of water level changes, but also provide a scientific basis for prediction. Next, a time series analysis algorithm including the autoregressive integrated moving average model (ARIMA) or the long short-term memory network (LSTM) is used to perform trend analysis on historical water level data.

[0052] These algorithms perform well when processing time series data. ARIMA is suitable for capturing linear relationships and autocorrelations of time series, while LSTM can effectively handle complex nonlinear relationships and long-term dependencies. By learning from historical data, these models can generate short-term and long-term water level forecasts to help relevant management departments formulate corresponding water resources management strategies. In addition, in order for the model to maintain efficient prediction capabilities under different environmental conditions, it is crucial to dynamically adjust the parameters of the spatiotemporal analysis model. This means that the parameters in the model need to be adjusted in a timely manner according to changes in external environmental factors such as seasonal precipitation and geological changes to ensure that the model can accurately reflect the current hydrological conditions.

[0053] The groundwater level anomaly monitoring and adjustment mechanism based on dynamic feedback first sets the dynamic threshold through the water level data predicted by spatiotemporal analysis. This is a key step because the traditional fixed threshold often cannot accurately reflect the actual water level changes. On this basis, the dynamic threshold can be automatically adjusted according to historical data and real-time monitoring information, thereby improving the sensitivity and accuracy of monitoring. Then, the system continuously compares the real-time monitoring data with the predicted water level value. Through this comparison, abnormal fluctuations in the water level can be quickly detected.

[0054] If the real-time data exceeds the dynamic threshold, the system will immediately identify this abnormal situation and trigger the subsequent adjustment mechanism. Another important part of the mechanism is to use feedback control algorithms to make immediate adjustments to abnormal situations. Feedback control algorithms can process monitoring data in real time and respond to detected abnormal situations to ensure the flexibility and adaptability of the monitoring system. When abnormal fluctuations occur, the algorithm can automatically adjust the monitoring frequency and scope, for example, increasing the monitoring intensity of abnormal areas to obtain more detailed information. This flexible adjustment capability ensures that sufficient data can be obtained to support decision-making at critical moments, and can also provide a more accurate basis for subsequent water resources management.

[0055] Embodiment 1:

[0056] Combination Figure 2 In a water resource management area, researchers used the optimization method of groundwater level dynamic monitoring to effectively manage water levels. In order to explain the whole process in detail, water level data and related environmental variables such as precipitation and temperature were first collected over the past year. The historical water level data are set as follows (unit: meter):

[0057]

[0058]

[0059] Next, construct the input data matrix X(t) and set the following parameters:

[0060] The water level data is X(t) = [5.2, 5.4, 5.1, 5.3, 5.5, 5.6, 5.8, 5.7, 5.9, 6.0, 5.5, 5.3]

[0061] The model parameters θ are initialized in the range of 0.01 to 0.1

[0062] The time weight function α(t′) is set to:

[0063] The weight from January to June is 1.0

[0064] July and August have weights of 1.5 and 1.2 (due to higher precipitation)

[0065] The weight from September to December is 0.8

[0066] When training the LSTM model, you first need to define the dynamic learning function g(τ,β). Set β = 0.5, and the calculation of this function can be performed as follows:

[0067] g(τ,β)=β·e -τ

[0068] For each time step t′, we can calculate For example, assuming calculations from January to July:

[0069]

[0070] This integral can be calculated by:

[0071]

[0072] Therefore, the water level forecast for each month can be calculated. Next, use the LSTM model for prediction. Set the future time step k = 3 months and calculate the prediction result:

[0073]

[0074] Specific calculation of water level forecast for the next 3 months:

[0075]

[0076] Assume that for the water level in December, the current water level is set to X(12) = 5.3. The water level after dynamic adjustment in the next three months is:

[0077] December: Forecast α(12) = 0.8

[0078] January: Forecast α(1) = 1.0

[0079] February: Forecast α(2) = 1.0

[0080] Through LSTM and dynamic adjustment model, it is assumed that the final prediction value is:

[0081]

[0082] Finally, the periodic characteristics of the water level data are analyzed by Fourier transform, and the calculation method of the periodic characteristics is set as:

[0083]

[0084] Here, N is the number of observation points, which is obtained by calculation:

[0085] Short-term trend: water level rises 0.3 meters per quarter

[0086] Long-term trend: water level rises by an average of 0.5 meters per year

[0087] Through these detailed calculations and data analysis, researchers can not only effectively predict water level changes in the next few months, but also combine real-time monitoring and feedback mechanisms to provide scientific basis and decision-making support for regional water resources management.

[0088] The construction of spatial coupling model is to combine geographic information system (GIS) and spatial statistical methods;

[0089] First, water level data from multiple monitoring points were collected, including the following monitoring points and their historical water level records (unit: meter):

[0090] Monitoring Points January April July October A 5.2 5.3 5.8 6 B 5.1 5.4 5.7 5.9 C 5.3 5.5 5.9 5.8 D 5.4 5.6 5.5 5.7

[0091] Next, the team constructed a spatial coupling model to quantify the degree of coupling of water level changes in different regions through spatial autocorrelation analysis. The Moran's I index was used to analyze the spatial correlation between monitoring points, and the number of monitoring points was set to N = 4. Assuming that the calculated Moran's I value is 0.4, it means that there is a moderate positive correlation between the monitoring points, which means that the water level change trends in some areas are similar.

[0092] Subsequently, the team built a multi-objective optimization model to optimize the scheduling strategy of groundwater resources. The specific optimization goal is to minimize the difference between water resource extraction and demand while controlling the scheduling cost. The following formula is used:

[0093]

[0094] Here, we set λ1 = 1.0 and λ2 = 0.5, which means that the team hopes to optimize in a way that balances extraction demand and cost. The comprehensive cost function C(t) includes energy consumption and resource loss costs, and the specific value is set to C(t) = 0.2, which represents the extraction cost per unit of water. The spatial weight matrix w ij The elements are defined as:

[0095]

[0096] For example, assuming that monitoring points A and B are adjacent, then w AB = 1. Next, the impact of water level changes is calculated using the measured water level values. Assuming that the water levels at adjacent monitoring points A and B are 5.8 meters and 5.7 meters respectively, substitute the formula:

[0097] X i -X j =5.8-5.7=0.1

[0098] Therefore, the square of the spatial coupling effect is calculated:

[0099]

[0100] The comprehensive cost is calculated as:

[0101]

[0102] Substituting these values ​​into the optimization formula, the researchers calculated the optimal extraction volume Q(t) and demand volume D(t) at time t through numerical integration, assuming that the extraction volume is 5.9 meters and the demand volume is 5.8 meters. The final objective function is calculated as:

[0103] λ1(Q(t)-D(t)) 2 +λ2·0.21=1.0·(5.9-5.8) 2 +0.5 0.21 = 0.01 + 0.105 = 0.115

[0104] Through this optimization process, the research team was able to effectively identify the regional characteristics of water level changes and adjust extraction strategies to achieve sustainable use of resources.

[0105] A feedback control mechanism is introduced to optimize the water resource scheduling strategy. First, the real-time water level data and historical data of multiple monitoring points are collected so that the changes in groundwater levels can be reflected in a timely manner. For example, suppose the water levels of monitoring points E and F at a certain moment are 6.0 meters and 5.9 meters respectively, while in the historical data, the water demand in the area is 500 cubic meters and the extraction volume is 480 cubic meters.

[0106] In this study, the team set control targets and used the objective function J(Q,D,θ) for comprehensive evaluation, where Q represents the extraction volume, D represents the water resource demand, and θ is the scheduling model parameter. In order to establish the scheduling model, it is necessary to calculate the partial derivatives of the objective function to identify the impact of different variables on the overall scheduling strategy. Assume that the calculated partial derivative is:

[0107]

[0108] Here, 0.02, -0.01, and 0.005 represent the influence of the extraction, demand, and scheduling parameters on the objective function, respectively. Next, the learning rate γ is set to 0.1, which indicates the speed at which the parameters are updated in each adjustment.

[0109] The team used the feedback control formula to adjust the parameters. The calculation formula is:

[0110]

[0111] Substituting the above partial derivative values, we get:

[0112]

[0113] This result indicates that at time t, the adjustment rate of the scheduling model parameters is 0.0015, which means that the model parameters will increase by 0.0015 after each feedback to better adapt to the current water level changes.

[0114] Next, the team dynamically adjusted the extraction and recharge strategy based on real-time monitoring data. Assume that in the new monitoring cycle, the real-time water level monitoring shows that the water level at point E has dropped to 5.8 meters, while the demand has increased to 550 cubic meters. According to the feedback control mechanism, the team needs to re-evaluate the extraction volume to meet the new demand. Assume that the current extraction volume is still 480 cubic meters and the model parameter has been adjusted to 0.005 (based on the previous adjustment).

[0115] The new objective function is:

[0116] J(Q,D,θ)=(QD) 2 +θ 2

[0117] Substituting the current data, we get:

[0118] J(480,550,0.005)

[0119] =(480-550) 2 +(0.005) 2 =(-70) 2 +0.00000025

[0120] =4900+0.00000025≈4900

[0121] This value indicates that the current scheduling strategy needs to be optimized. To this end, the team generated a new scheduling strategy through genetic algorithms, taking into account multiple objectives of environmental changes and resource allocation. By simulating different extraction volumes and demand volumes, and through iterative optimization, the final extraction volume was determined to be 510 cubic meters.

[0122] Embodiment 2:

[0123] Combination Figure 3 In a water resources management project, the research team was committed to optimizing the abnormal fluctuation detection method of groundwater level dynamic monitoring through spatiotemporal analysis. To achieve this goal, the historical water level data of each monitoring point was first collected and the relevant environmental variables such as precipitation and evaporation were recorded.

[0124] For example, in the past month, the historical water level data of monitoring point A was: 5.5 meters, 5.6 meters, 5.4 meters, 5.7 meters, 5.6 meters, and the precipitation was: 10mm, 15mm, 5mm, 20mm, 0mm, and the corresponding evaporation was: 3mm, 4mm, 2mm, 5mm, 1mm. By analyzing these data, the team calculated the mean μ(W(t)) and standard deviation σ(W(t)) of the historical water level data and obtained:

[0125]

[0126] Next, the team developed a dynamic threshold formula:

[0127]

[0128] In this formula, the team chose a control coefficient k of 0.5 to increase the sensitivity of the dynamic threshold. To consider the impact of periodic fluctuations, α = 0.2, P = 30 (analysis period is 30 days), φ = 0 (the starting position of the fluctuation), β = 0.1 and γ = 0.05 were set. Then, the precipitation R (t) = 15 mm and the evaporation E (t) = 4 mm on a certain day (for example, the 10th day) were calculated.

[0129] Substituting these values ​​into the dynamic threshold formula, we obtain:

[0130]

[0131] First calculate the sine part:

[0132]

[0133] Then bring it into the calculation of dynamic threshold:

[0134] T(10)=5.56+0.5·0.079·(1+0.2·0.866+1.5+0.2)

[0135] =5.56+0.5·0.079·(1+0.1732+1.5+0.2)

[0136] =5.56+0.5·0.079·2.8732

[0137] ≈5.56+0.5·0.226

[0138] ≈5.56+0.113≈5.673 meters

[0139] After calculating the dynamic threshold T(10)≈5.673 meters, the team compared the real-time monitoring data with this threshold. When the actual water level monitored is lower or higher than this threshold, the system automatically issues an alarm to indicate abnormal fluctuations. For example, if the actual water level drops to 5.5 meters, the system will recognize this fluctuation and trigger further analysis.

[0140] Based on previous work, the dynamic threshold T(t)≈5.673 meters has been calculated, and the water level changes at each time point have been monitored in real time. On this basis, the team further introduced a real-time comparison mechanism to continuously compare the current water level data with the predicted dynamic threshold to identify abnormal fluctuations and trigger alarms.

[0141] Assume that on a certain day (the 11th day), the actual monitored water level W real (t) is 5.4 meters, and the team set the tolerance range ∈ to 0.1 meters to reduce the possibility of false alarms. The tolerance range is set to avoid triggering unnecessary alarms within the normal fluctuation range while ensuring sensitivity to larger anomalies.

[0142] Use the alert formula:

[0143]

[0144] Substituting the values, the dynamic threshold T(10) = 5.673 meters, plus and minus the tolerance range are:

[0145] T(t)+∈=5.673+0.1=5.773 m

[0146] T(t)-∈=5.673-0.1=5.573 meters

[0147] Real-time water level W real (t) = 5.4 meters. According to the formula:

[0148] iW real (t) <T(t)-∈,thenAlert=1

[0149] since5.4<5.573,thusAlert=1

[0150] The system identified abnormal fluctuations in water levels and triggered an alarm. The research team immediately took corresponding measures, including increasing the frequency of monitoring in the area and strengthening the collection of water level change data to ensure that the dynamic changes in water levels are captured more accurately. At the same time, the team analyzed the possible causes of abnormally low water levels, such as reduced precipitation, increased extraction, or geological changes, and adjusted the recharge strategy in a timely manner to replenish groundwater resources. Through this real-time feedback and dynamic adjustment mechanism, the team was able to respond quickly to water level anomalies and prevent a wider range of water resource crises.

[0151] In addition, in order to reduce the waste of resources caused by false alarms, the team optimized the setting of tolerance ∈ and found that when ∈ was set to 0.05 meters, the alarm frequency increased significantly, but in most cases did not bring serious consequences. Therefore, the relatively loose ∈ = 0.1 meters was finally selected to balance sensitivity and stability.

[0152] The implementation steps include: first, uploading all monitoring data to the central processing system in real time, and the system continuously calculates the current dynamic threshold and performs real-time comparison; second, based on the comparison results, the system determines whether to trigger an alarm; if triggered, an alarm message is immediately sent to the water resources management department, and it is recommended to take countermeasures, such as adjusting the operating parameters of the pumping equipment, starting backup water supply, etc.; finally, the causes of abnormal fluctuations are analyzed, and the setting of alarm thresholds is continuously optimized through the data feedback mechanism to ensure that the monitoring system can maintain efficient response under different environmental conditions.

[0153] In the further implementation of the water resources management project, the research team began to combine time series prediction with spatial coupling analysis to identify the correlation between different monitoring points and improve the accuracy of overall water level prediction in the region. The team used historical water level data, environmental variables (such as precipitation and evaporation), and spatial coupling data to build a deep learning model to predict future water level changes by weighting these historical data.

[0154] Specifically, the recent data collected by the team at monitoring point A are: water level W(t) is 5.6 meters, precipitation R(t) is 20 mm, and evaporation E(t) is 4 mm. In addition, spatial coupling related data S(t′) was also collected, such as the water levels of adjacent monitoring points B, C, and D are 5.8 meters, 5.5 meters, and 5.7 meters, respectively. Based on these data, the team predicts future water levels through the mapping function g(·) of the deep learning model. The mapping function integrates all relevant factors and performs weighted averaging to comprehensively predict water level changes after the next k time steps.

[0155] The calculation formula for the predicted water level is:

[0156]

[0157] In the model, the team set the range of values ​​for each parameter: α = 0.4 (weight of water level), β = 0.3 (weight of precipitation), γ = 0.2 (weight of evaporation), δ = 0.1 (weight of spatial coupling). These weights are set based on the analysis of historical data and the results obtained through deep learning model training. The model parameter set θ contains the weights and biases of the network. These parameters have been optimized through multiple iterations of training, so that the model can fully capture the complex relationship between various influencing factors. Substituting the above parameters and data for prediction, assuming k = 3 (i.e. predicting the water level changes in the next three days), the integral calculation process is:

[0158]

[0159] Compute the weighted average of the parts:

[0160]

[0161] Integrate this result:

[0162]

[0163] The predicted water level change results show that the water level in the area may rise to 28.821 meters in the next three days. This prediction result was compared and verified with the data of a single monitoring point, showing consistency with the change trend of other points, indicating the effectiveness of the model in spatial coupling analysis.

[0164] The team then applied the forecast results to actual scheduling, and further refined water resource management measures by comparing the predicted future water level values ​​with the current dynamic thresholds. For example, assuming that the target water level in the area is maintained within 6.0 meters, when the forecast results show that the future water level will exceed this threshold, the team will adjust the groundwater extraction strategy in advance, reduce the extraction volume or increase the supply to stabilize the water level.

[0165] The specific implementation steps include: the first step is to collect real-time data from each monitoring point and input it into the model; the second step is to use the deep learning model to make time series predictions and combine the spatial coupling factors for weighted calculations; the third step is to compare the prediction results with the set target water level range to identify possible abnormal fluctuations; the fourth step is to adjust the water resource scheduling strategy based on the prediction results and spatial analysis to avoid excessive extraction or recharge. Through this complete implementation process, the team has effectively improved the management efficiency of groundwater resources, ensured that the water level fluctuates within a reasonable range, and prevented potential water resource crises.

Claims

1. The optimization method of groundwater level dynamic monitoring is characterized by The following steps are involved: S1. Groundwater level monitoring based on sensor array: S1.

1. Multiple sensors are evenly distributed in the groundwater monitoring area, including water pressure, temperature, humidity and other multi-dimensional parameter monitoring; S1.2, the sensors form a grid monitoring system through a wireless network; S2. Multi-parameter collaborative groundwater level data fusion: S2.1, input the multi-dimensional data collected by the sensor array into the data fusion module, use the data fusion algorithm to coordinate the data and filter out noise and outliers; S2.2, build a unified water level change model so that data of different parameters can complement each other; S2.3, regularly adjust the weights in the fusion algorithm based on historical data and new measurements; S3. Prediction of groundwater level change trend based on spatiotemporal analysis: S3.

1. In the monitoring area, based on the dimensions of time and space, a spatiotemporal analysis model of groundwater level changes is constructed to explore the periodicity and regularity of water level fluctuations; S3.2, use time series analysis algorithms including ARIMA or LSTM to perform trend analysis on historical water level data and generate short-term and long-term water level forecasts; S3.

3. Dynamically adjust the parameters of the spatiotemporal analysis model to adapt to changes in different environmental factors including seasonal precipitation and geological changes; S4. Groundwater level anomaly monitoring and adjustment mechanism based on dynamic feedback: S4.

1. Set dynamic thresholds based on water level data predicted by spatiotemporal analysis, monitor and compare actual data with predicted values ​​in real time, and detect abnormal fluctuations; S4.

2. Use feedback control algorithms to make immediate adjustments to abnormal situations and increase monitoring efforts in abnormal areas by adjusting monitoring frequency and range; The method for constructing the spatiotemporal analysis model includes: using a long short-term memory network (LSTM) to train historical water level data and environmental variables, capturing the nonlinear dynamic characteristics of time series, and predicting future water level changes; combining Fourier transform to extract periodic characteristics, identify short-term fluctuations and long-term trends in water level changes; and exploring the inherent laws of water level changes: in, is the predicted future water level change value, which represents the water level prediction result after k time steps in the future; X(t) is the input data matrix at the current moment, which contains the historical water level data of all monitoring points; θ is the parameter set of the LSTM model, including weights and biases; α(t ′ ) represents the time weight function of the input feature, reflecting the importance of the feature at different time points; t ′ is the time step; g(τ,β) is the time-dependent dynamic learning function, which describes the time dependence of water level change, and the parameter β controls the shape and sensitivity of the function; The nonlinear trend of water level change is captured by internal integration of the function g(τ,β); the change of water level within k time steps is comprehensively predicted by external integration; Fourier transform is used to transform the water level change in time domain into frequency domain analysis, identify the periodic characteristics P of water level fluctuations, and form a comprehensive description of short-term and long-term trends.

2. The optimization method for dynamic monitoring of groundwater level according to claim 1, characterized in that The method for constructing the spatiotemporal analysis model includes: constructing a spatial coupling model, performing spatial correlation analysis on water level data of different monitoring points by combining a geographic information system (GIS) and a spatial statistical method, quantifying the coupling degree of water level changes in different regions, and identifying regions of related or independent changes; and optimizing the scheduling strategy of groundwater resources by constructing a multi-objective optimization model.

3. The optimization method for dynamic monitoring of groundwater level according to claim 2 is characterized in that The method for constructing the spatiotemporal analysis model includes: introducing a feedback control mechanism, using real-time monitoring data and prediction results to dynamically adjust the extraction and replenishment strategies; combining genetic algorithm optimization control strategies to control the dynamic balance of multiple objectives, and the adaptive control module adjusts the control parameters to form a closed-loop optimization system when receiving new water level change information, feedback control and optimization functions: in, is the adjustment rate of the scheduling model parameters, indicating the speed of parameter change over time; Q represents the amount of groundwater extracted; D is the demand for water resources; J(Q, D, θ) is the objective function, which comprehensively considers the relationship between extraction, demand and control parameters, and reflects the degree of optimization of the current scheduling strategy; are the partial derivatives of the objective function with respect to the extraction quantity, demand quantity and parameters, respectively, which measure the impact of the variables on the scheduling strategy; γ is the learning rate, which controls the adjustment speed.

4. The optimization method for dynamic monitoring of groundwater level according to claim 1 is characterized in that The method for detecting abnormal fluctuations includes: dynamically calculating the threshold of each monitoring point through spatiotemporal analysis, combining historical water level data and environmental variables including precipitation and evaporation; the dynamic threshold reflects the fluctuation of the current water level and is adjusted according to seasonal changes, climate conditions and historical fluctuation patterns.

5. The optimization method for dynamic monitoring of groundwater level according to claim 4 is characterized in that The method for detecting abnormal fluctuations includes: continuously comparing real-time water level data with predicted dynamic thresholds; identifying abnormal fluctuations in water levels, triggering alarms and taking corresponding measures; and reducing false alarms by setting a tolerance range: Among them, Alert indicates the alarm state. When the real-time water level W real When (t) exceeds the dynamic threshold T(t) plus the tolerance ∈ or minus the tolerance ∈, the alarm value is 1, indicating that an alarm is triggered; otherwise it is 0, indicating a normal state.

6. The optimization method for dynamic monitoring of groundwater level according to claim 1 is characterized in that The method for detecting abnormal fluctuations includes: identifying the correlation between different monitoring points by performing water level time series prediction and combining it with spatial coupling analysis; based on the data of a single monitoring point, covering the water level changes at other points in the area; and predicting the results by weighting historical data.

Images